DIY '2400cc Inference Racer': Dual RTX 3090s, ThinkPad ECU and a VW Golf Radiator Serve 27B Models
TooManyPascals · reddit · 2026-09-26
A Redditor unveiled the "2400cc Inference Racer", a home inference rig doubling as a winter heater:
- Hardware: two used AORUS RTX 3090 XTREME WATERFORCE cards linked via NVLink, hosted by a naked Lenovo ThinkPad motherboard (Ryzen 7 7840U, 64GB RAM) mounted with a custom water block and a clamp.
- Cooling: a €24 VW Golf radiator with mismatched hoses; 2.4L of coolant (hence 2400cc), leaking under 100ml/day.
- PCIe: one GPU on the WWAN slot (Gen4 x1), one on the SSD slot (Gen2 x4); BIOS patched to remove whitelist, alter power-up sequence and disable PCIe power-saving.
- Workload: Ubuntu + vLLM serving Qwen3 27B at surprisingly decent speeds; without the 3090s it acts as a low-idle server running small models via Vulkan/llama.cpp. No hot-swap yet.
More from Infra
- Running Qwen2.5-Coder-3B on Android: Jupyter + Pydantic Evals on-device eval pipeline — alex_b3b · 2026-11-11
- 22% of Nvidia revenue now books to Singapore, but only ~1% is local buyers — Gregory_C_Allen · 2026-11-11
- Huawei's Kirin 9050 NPU claims offline 30B MoE on smartphones, +140% perf — AIFlow_ML · 2026-11-11
- Stratechery: Intel vs AMD's diverging earnings and Intel's need for an old-new path — Stratechery · 2026-11-11
- a16z Podcast: Mike Volpi on What the 1990s Internet Boom Teaches Us About the AI Buildout — a16z Podcast · 2026-11-11
- Alibaba files to list chip-design unit T-Head in Hong Kong to fund AI push — pstAsiatech · 2026-11-11